On Wed, Apr 21, 2010 at 1:59 AM, A B <gentosaker(at)gmail(dot)com> wrote:
>> Sometimes cancel_backend fails because there's no point where the code
>> checks for a cancel.
>
> What would such a point in the code look like?
For instance, if your backend is waiting on a socket call in libc, as
in this thread:
http://archives.postgresql.org/pgsql-general/2010-03/msg00663.php
Josh